010010001000 01001000100000110000001000001100 DLP Driven, Learning, Optical Neural Networks Emmett...
-
Upload
stewart-barton -
Category
Documents
-
view
213 -
download
1
Transcript of 010010001000 01001000100000110000001000001100 DLP Driven, Learning, Optical Neural Networks Emmett...
01001000100001001000100000110000001000001100
DLP Driven, Learning, Optical Neural Networks
Emmett Redd &
A. Steven Younger
Associate Professor
Missouri State [email protected]
2
Neural Network Applications
Stock Prediction: Currency, Bonds, S&P 500, Natural Gas
Business: Direct mail, Credit Scoring, Appraisal, Summoning Juries
Medical: Breast Cancer, Heart Attack Diagnosis, ER Test Ordering
Sports: Horse and Dog Racing
Science: Solar Flares, Protein Sequencing, Mosquito ID, Weather
Manufacturing: Welding Quality, Plastics or Concrete Testing
Pattern Recognition: Speech, Article Class., Chem. Drawings
No Optical Applications—We are starting with Boolean
most from www.calsci.com/Applications.html
3
Optical Computing & Neural Networks
Optical Parallel Processing Gives SpeedLenslet’s Enlight 256—8 Giga Multiply and
Accumulate per secondOrder 1011 connections per second possible with
holographic attenuators
Neural NetworksParallel versus SerialLearn versus ProgramSolutions beyond ProgrammingDeal with Ambiguous InputsSolve Non-Linear problemsThinking versus Constrained Results
4
Optical Neural Networks
Sources are modulated light beams (pulse or amplitude)
Synaptic Multiplications are due to attenuation of light passing through an optical medium
Geometric or Holographic Target neurons sum signals from many source
neurons. Squashing by operational-amps or nonlinear
optics
5
Standard Neural Net Learning We use a Training or Learning algorithm to adjust
the weights, usually in an iterative manner.
… x1
xN
y1
…yM
…
Learning Algorithm
Target Output (T) Other Info
6
FWL-NN is equivalent to a standard Neural Network + Learning Algorithm
FWL-NN
+Learning Algorithm
7
Optical Recurrent Neural Network
Squashing Functions
Synaptic Medium(35mm Slide)
Target Neuron Summation
Signal Source (Layer Input)
Layer OutputA Single Layer of an Optical Recurrent Neural Network. Only four synapses are shown. Actual networks will have a large number of synapses. A multi-layer network has several consecutive layers.
Micromirror Array
PresynapticOptics
Postsynaptic Optics
Recurrent Connections
8
Optical Fixed-Weight Learning Synapse
W(t-1)
x(t)
Σ
x(t-1)
T(t-1)
Planapse
Tranapse
y(t-1)
9
Definitions
Fixed-Weight Learning Neural Network (FWL-NN) – A recurrent network that learns without changing synaptic weights
Potency – A weight signal Tranapse – A Potency modulated synapse Planapse – Supplies Potency error signal Recurron – A recurrent neuron Recurral Network – A network of Recurrons
10
Page Representation of a Recurron
Tranapse
Planapse
Tranapse
Tranapse
U(t-1)
A(t)
B(t)
Bias
O(t-1)Σ
11
Optical Neural Network Constraints
Finite Range Unipolar Signals [0,+1] Finite Range Bipolar Attenuation[-1,+1] Excitatory/Inhibitory handled separately Limited Resolution Signal Limited Resolution Synaptic Weights Alignment and Calibration Issues
12
Proposed Design
Digital Micromirror Device 35 mm slide Synaptic Media CCD Camera Synaptic Weights - Positionally Encoded
- Digital Attenuation Planapse, Tranapse trained using Standard
Backpropagation Allows flexibility for evaluation.
13
Generation, Acquisition, Processing
DemonstrationLeft Images (time order)
• Gray-scale Source Neuron Signal• Equivalent Pulse Modulated Source Neuron Signal• Gray-scale Source Neuron Signal – Reprise• Blank during Target Summation and Squashing
Middle Image – Fixed Weights on Attenuator Slide (static)
Right Images (time order)• Blank during First Source Gray-scale• Accumulating (Integrating Pulses) Target Neuron Input• Gray-scale during Summation and Squashing
Phases and Layers• Synapses Connect Two Layers (N and N+1)• Only Source Neurons (layer N) Illuminate• Only Target Neurons (layer N+1) Accumulate
14
Hard and Soft Optical Alignment
Hardware AlignmentBest for Gross AlignmentPrecision Stages Required for Precise AlignmentHard to Correct for Distortions in Optics
Software AlignmentFine Image Displacements Image Rotation Integrating Regions of Arbitrary Shape and SizeDiffraction Effects Filtered and IgnoredCorrect for Optical Path Changes
15
Photonic Circuit Elements
NEAROptical FiberFiber AmplifiersSplitters
MEDIUMCylindrical Lenses before and after Synaptic
Media (1-D systems, sparse media)Non-Linear Crystals for Squashing
FAR Integrated Emitters, Attenuator, Detectors/LimitersHolographic Attenuators (destructive interference)Stackable
16
DLP Driven, Learning, Optical Neural Networks
Emmett Redd and A. Steven Younger
Associate ProfessorMissouri State [email protected]
This material is based upon work supported by the National Science Foundation under Grant No. 0446660.